Digital camera basics
The digital camera truly relects the age in which we live. That is not a compliment to either. Development funds have been poured into making sure
any half-wit can take a half-decent digital photo, and secondly
no-one else can take anything better.
The problem is not any evil giant corporation, but a common substance commonly found in bulk beneath my bed - dust. An imaging chip is small, to save silicon, so any dust specks on it appear large in the finished picture. If you can remove a camera lens, in flies the dust. The camera lens is thus non-removable, so has to be power-zoomed to fill the frame with the subject. No-one can see fine detail on the little screen on the back, so auto-focus completes the package. Bingo - the half-decent digital photo. The professional digital cameras have a removable lens and manual focus, at the cost of staggering complexity. Every so often, at 100 US dollars a time, the owners return them to the makers to have the dust cleaned out of the inside. Slowly a really sensible and fully professional answer to the problem is emerging. The imaging chips are being made with a frame size of 24 x 36 mm, or the normal 35 film standard. On this giant chip, dust specks are tiny, and the problem goes away. Even if a professional camera were affordable, since the maximum chip area a microscope objective will illuminate is only about 6mm x 5mm, most of the big chip area is wasted.
Higher maths - pixels and resolution
Resolution, both of lenses and imaging media (including film and silicon chips) is traditionally measured in lines per millimetre or lpmm for short. A target made of black lines on a white ground is photographed. The white spaces between the parallel black lines are the same width as the black lines, so the target could equally well be described as white lines on a black ground. The target is progressively distanced from the test camera, so the created image gets smaller. The smallest image on which the black/white line pattern can still be discerned is checked on a measuring microscope to see how many parallel black lines fit in 1mm of image. If the film can resolve more than the lens, you are testing the lens. If the lens outperforms the film, you are testing the film. For a long time now a 'good' lens has resolved at least 70 lpmm, and a high resolution film about 90 lpmm. Thick colour films like Kodak Gold are around 60 lpmm.
A good photographic rule of thumb has been that photography to 70 lpmm is quite easily done. Anything better than the magic 100 lpmm is virtually impossible. Let us do some maths with a very common silicon imaging chip.
An imaging chip contains pixels, lying in parallel rows to create a rectangular array. An array roughly 3mm by 4mm can be found on the VGA sensor, used in all quality webcams. The pixel dimensions on the chip are 480 pixels high, and 640 pixels wide. The total number of pixels is thus
480 x 640 = 307200
This chip is thus described as the 300,000 pixel VGA chip, and if you look at webcam specs, you will see this quoted.
480 pixels in 3mm equals 160 pixels in 1mm, so you have 160 lines of pixels in the imaging chip per millimetre. You need a pixel line to hold the black line of the resolution test, and the pixel line beside it to hold the white separator. The maximum resolution of this chip is thus 160 divided by 2, or 80 lpmm. Plainly the chip designers read up on photography before starting. Most imaging chips are designed around this resolution. The Sony chip used in their base-model video cameras is specified as a 1/6 inch diameter sensor for 800,000 pixels. Most of their competitors use 1/5 inch or 1/4.5 inch for 800,000 pixels. If you calculate it out the Sony chip must have a resolution of well over 100 lpmm. They either have some marvellous lenses to match this resolution, or are simply scrimping on silicon. No comment.
If you check the 'background' section you will find photomicrographs taken on a good quality 300,000 pixel VGA webcam chip from Kodak, and they are pretty good, although the basic pixels of the images are not far below the surface. Any microscope objective covers 3mm by 4mm with room to spare. The next step up is the '1.3 Megapixel' chip, which was the ultimate in digital cameras a couple of years ago. If you work out the figures for a 6mm by 8mm chip, with 960 lines each holding 1280 pixels, you get 1,228,800 pixels. The same design rules have been followed as for the VGA chip, and the resolution is again 80 lpmm. A good Plan objective will image 6mm x 8mm right to the corners, so the classic '1.3 Megapixel' chip is perfect for photomicrography. A larger chip size than this is too big for the objective field, and the objective itself is pushing to resolve 80 lpmm, so adding more pixels to this size chip will not get you any more sharpness. For now, and for ever, a 1.3 Megapixel 6mm by 8mm chip is all that you need. As the years pass, until production of this elementary size stops, there will be slow improvements in colour quality and contrast handling.
The clean mathematics above is sullied by two games played by the digital camera makers. The first obstructs good photomicrography, but the second makes precise focusing possible. Both techniques concern manipulation of the pixels by the camera microprocessor after the image has been sucked off the sensor chip into the circuit board.
It is possible for the camera microprocessor to examine each side-by-side pair of pixels in an image, and make a good guess at what value a third pixel would hold, if it were located in between the pair. This value is of course the averaged value of the two real pixels round it. A whole image may then be built up containing generally four times the number of pixels as the original. At first view such an image is much less 'blocky', and more pleasing to the eye, but the interpolation process cannot add any more detail, and the result adds fuzziness to a photomicrograph, which helps nobody. If you really want to 'resample' the picture to more (or less) pixels, you can do it in the computer, as any photo-manipulation package has this facility.
For a long time, the VGA 300,000 chip was all cheap digital cameras could afford. Some makers incorporated interpolation into their product, generating a picture with four times the pixels of the sensor. The English language was subtly manipulated to suggest to as many customers as possible that the camera contained a 1.3 Megapixel chip, without actually saying so. The best example I found was Pixera, who made, and may still be making, a 5000 US dollar microscope camera containg a VGA 300,000 chip. The top line of their sales sheet says "1.3 million real pixels" in heavy type. At the end of the third page of the data sheet is a tiny line saying the pixels are interpolated from a VGA chip. D-Link sold quite a few VGA cameras as 1.3 Megapixel, but were a little more honest with the wording.
The first really cheap camera to contain a genuine 1.3 Megapixel imaging chip was used, of course, as the basis for our current microscope camera.
A 2X digital zoom on a 1.3 Megapixel chip simply discards all but the centre 300,000 pixels, and blows these up to fill the viewing screen on the back of the camera giving a pseudo-telephoto effect. When such a picture is 'snapped' an interpolation (see above) is performed on these centre pixels to create a 'normal' 1.3 megapixel image of an enlarged subject. Nothing has been gained of course, and you might as well have left the digital zoom alone, and enlarged and resampled the centre of the picture in your computer if you wanted the same effect.
If you are focusing, as in photomicrography, using the LCD screen on the back of the camera, such a screen has a low resolution, normally 240 by 320 pixels, and makes 'guessing' of correct focus difficult. If you digital zoom on the point to focus on, the subjectively greatly enlarged image makes focusing relatively easy. The zoom is switched off again before the picture is taken. The digital zoom adds nothing to the normal digital camera but a selling feature, but does make focusing in photomicrography possible.
Photographic film has undergone research (and development of course) for nearly 150 years. Solid-state image sensors are comparatively recent. The new device is much more subject to 'highlight burnout' where areas receiving too much light just reset to blank white. This does not wreck the sensor, but it does wreck your picture. The Kodak webcam sensor has a very basic auto-exposure that just progressively turns down the sensitivity as the light level rises. The 1.3 Megapixel circuit we use is plainly much more of a hair-trigger device, and reacts intelligently to the inclusion in the picture of small bright areas. Basic beliefs have us wishing we could just turn off auto-exposure, but on balance it probably does more good than harm. On expensive digital cameras, the lens has a power-driven iris aperture control, which can actually reduce the light level reaching the sensor. If you can control the illumination manually, with a lens iris, or by turning the microscope illumination down, you can keep the auto-exposure in a light range where it does no harm.
The Infra-Red menace
The effect of Infra-Red (IR) just outside the visible spectrum has to be looked at in depth. All silicon-based sensors are much more sensitive to IR than they are to visible light. Unless the IR is removed from the incoming light, the picture loses most of its colour, turning into a purplish monochrome. Since the imaging lens is not corrected for IR, and in addition focuses it at a different point, unless the lens is stopped right down, the picture is unsharp as well. Unless you are after black skies, black eyes and white trees the IR must be filtered out.
To totally exclude the IR and leave the visible, normal filters are not adequate. An interference filter is required, in which a thin layer of coating, of a precise refractive index and thickness, is applied to the glass filter blank. Unwanted IR destructively interferes with itself at the coating interface, and visible light passes through. According to the Internet, scientific interference filters start at about 600 US dollars each. Most are made in the US, but Schneider, the formidable European optical maker, has a good range. Reduced-specification IR filters for general photography are unheard-of in Australia but common in the US. These are known as 'hot mirror filters'. The largest Photo dealer in Melbourne was unaware of this product, but it was in the Canon product listing. I ordered in this US-made filter. Surface quality was atrocious, and it cost 90 Australian dollars, but it cut the IR out as required.
An early digital camera made by Kodak actually came with a hot mirror filter, but since that time, because the lens is a fixture, the filter coating was applied to the back element of the camera lens for the cost, at manufacture, of a few cents extra. As soon as the fixed lens is removed, as in our camera, IR sensitivity once more is an issue. Normal microscope illumination systems are awash with IR. You can either put a hot-mirror filter over the light where it passes into the condenser (we have done this successfully) or alternatively fit a ledlyt to the microscope. This provides ideal photographic (and visual) illumination, and there is no IR at all generated by the ledlyt. Saying it again, unless you plan to fit a lens, or lenses, to our microscope camera to use it for technical photography as well, if you have a ledlyt installed, there is no need for a hot mirror filter.
The camera/computer connection
Webcam/computer links are much harder to set up than camera/computer links, since the webcam is passing live video, which has to be displayed on a fairly modern system if you want a picture of reasonable size. USB is now the common connection to both, but all the camera needs to do is send the taken pictures down the wire to the computer. Various interfaces exist. The one we use is pretty common, in which the camera memory emulates an external hard disk. With the manufacturers software installed under Windows 98SE, normal drag-mouse type file extraction operated without trouble, but I had to get crafty when using DOS command-line techniques to download 130 pictures in one hit. The big surprise was running the camera and USB connection under Red Hat Linux 8. No extra software was needed to set it up. The internal Linux USB driver sufficed. Command-line bulk transfer of files was totally bug-free under Linux.